AI is no longer a novelty. What began as a wave of experimentation – with chatbots, pilots and proof-of-concepts – has now reached a new phase. Companies are under pressure to evolve, create value and show the tangible impact of AI beyond the demos. And many mid-size firms are finding this challenging.
Worryingly – New figures show that 42% of companies have now shelved the majority of their AI projects, up from 17% last year. While tech giants like Microsoft and Google continue to charge ahead, mid-market firms and SMEs risk being left further behind as the initial wave of AI enthusiasm has given way to scalability issues, infrastructure gaps and unclear strategy.
This is not a technology problem, but a business one. Companies must confront what really has to change to make AI work – it demands structural, strategic and behavioural change. Too often, businesses underestimate just how much has to move to make AI stick and scale.
Why AI initiatives stall
The first problem is lack of clear strategy. Businesses are eager to jump on the AI train, but fail to identify a clearly defined goal. Wanting to be able to talk about your businesses’ AI capabilities within its annual report is no longer enough – businesses must define what they mean by AI, what the tangible goals are and how this will change the business.
But the real issue isn’t a lack of interest, but a lack of preparedness. Mid-market businesses often assume that AI can simply be layered onto what already exists. But this is not the case – new capabilities demand new ways of doing things from top to bottom. Successful integration of AI shifts how mission-critical decisions are made, how work is distributed, and how humans interact with the business. This can’t be ignored. Nor can the fact that many firms are still operating on fragmented data and legacy platforms – even the best models will fail with these structural limitations.
Culturally, the challenge is just as big. AI is still widely misunderstood inside many organisations and across the workforce. As a result, it is often resisted. If staff aren’t educated and equipped to use these tools, and businesses don’t recognise how employees can collaborate with AI, then adoption will falter. The point is not whether AI will replace people, but that it will change the way they work. Those who don’t adapt risk being outpaced by their peers that do. This relies on communication from above, and support for employees must be baked into the initiative from day one.
These barriers combine to create a pattern that we’re seeing again and again. Namely, a burst of early-stage activity, followed by a slow fade out as organisations struggle to scale initiatives. This limits the value delivery, weakens the case for further investments, and means that AI slowly drops down the agenda.
How mid-market firms can win
Yet the mid-market is well-positioned to benefit when it comes to AI. With fewer silos and simpler governance, smaller organisations can move faster than their bigger competitors, provided they’re clear on where they are going and how they will get there.
One advantage is that foundational models are now commoditised. Mid-market firms should buy for parity and build for advantage, using off-the-shelf models for generic tasks, but retaining control over the data, workflows and decisions that differentiate your business.
Early wins for mid-sized firms can come from value-add activities that drive efficiency and cost savings, without requiring moonshot innovation. These include intelligent automation use-cases like customer service and AP invoicing. This is where mid-sized firms can outmanoeuvre their larger peers – they can move quickly, align cross-functional teams, and adapt without waiting for perfection.
But none of this works without leadership, and success requires a business transformation. The firms that succeed are those where executives engage directly with the tools, ask hard questions about ROI, and fund capability-building. They understand that, if they are to scale AI, the organisation must evolve with it.
Through my work I have advised hundreds of businesses on AI adoption, and one pattern is clear – the difference between experimentation and scale lies in how firms approach the fundamentals. The most successful initiatives start small but are focused on scale – piloting in real-world operational conditions and staying flexible enough to adjust and adapt. They scope carefully and make sure to define and track value early. They also treat imperfect data as a challenge to overcome, not a reason to stop. Critically, these firms link AI to their broader, strategic roadmaps, becoming system critical capability and accept that organisational change is needed. Without that, even the most promising AI pilots will never leave the sandbox.
